# charent/ChatLM-mini-Chinese

中文对话0.2B小模型（ChatLM-Chinese-0.2B），开源所有数据集来源、数据清洗、tokenizer训练、模型预训练、SFT指令微调、RLHF优化等流程的全部代码。支持下游任务sft微调，给出三元组信息抽取微调示例。

Repository: https://github.com/charent/ChatLM-mini-Chinese
Canonical: https://ross.abutalabs.com/products/chatlm-mini-chinese
Language: Python
License: Apache-2.0
License Family: permissive
Topics: chatbot, language-model, t5-model, text-generation
Last push: 2024-04-20T00:38:49+00:00

## Health v2 (maintenance only)
Score: 28/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 78
- inputs: {"age_days": 1102, "days_push": 866, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1728, forks 192 (observed 2026-08-28T04:05:28.319586+00:00)

## What it is
ChatLM-Chinese-0.2B is a small (0.2B parameter) Chinese conversational language model with fully open-sourced training pipeline code covering data cleaning, tokenizer training, pretraining, SFT instruction tuning, and DPO/RLHF preference optimization. It is built on the HuggingFace ecosystem and includes a fine-tuning example for triple information extraction.

## Use cases
- train a small chinese chat model from scratch
- learn the full llm training pipeline including pretraining sft and rlhf
- fine-tune a small model for triple information extraction
- run a chat model on low vram consumer hardware
- preprocess and deduplicate large chinese text datasets
- experiment with dpo preference optimization using lora

## When to choose
- you want to learn or reproduce the end-to-end process of training a language model from scratch
- you need a tiny chinese chat model that runs on 4GB vram or less
- you want open reference code for data cleaning, tokenizer training, SFT, and DPO
- you need a base for fine-tuning small downstream NLP tasks like information extraction

## When to avoid
- you need state-of-the-art chinese language understanding or generation quality
- you want a production-ready chatbot with strong reasoning abilities
- you need multilingual or english-focused capabilities
- you want a plug-and-play model without touching training code

## Facets
- artifact type: learning-resource
- maturity: maintenance
- function: llm-training, llm-inference, chatbot, nlp, machine-learning
- domain: large-language-models, machine-learning, chatbots, education
- platform: python
- tags: chinese-language-model, small-language-model, pretraining, sft, dpo, rlhf, t5, huggingface, information-extraction, from-scratch-training, natural-language-processing, gpu, linux

## Member repositories
- charent/ChatLM-mini-Chinese (main) score 28

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:28.319586+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:32:08.340081+00:00, confidence not recorded.
  - readme: https://github.com/charent/ChatLM-mini-Chinese (fetched 2026-08-28T04:05:28.319586+00:00, sha fb1e91292d05)
- Data as of 2026-08-30T08:39:29.467469+00:00.